Prime95 stability problems/questions

[zbose]

First my system:
Barton 3000+
1 GB Mushkin 222 special (2x512) (running 2.8v @ 2-2-2-11)
Asus A7N8X deluxe, rev 2

So here is the story. About two weeks ago, I got my 222 special. I found that I couldn't get it to pass the Prime95 torture test (third one, most memory tested) unless I upped the voltage to 2.8. At that point it ran fine. I called Mushkin's tech support and they assured me that they still warranty the ram if you run it at 2.8v (lifetime warranty). Great.

So, for my next upgrade, I decide to get some moderate watercooling (first time) and I bought a Hydrocool 200ex. My CPU has always run hot (up to 58C under load, usually 55C under load, 49 idle) but it never bugged my because my system seemed nice and stable. Well during the watercooling install, there was a slight problem... I tried using one of the fan only power lines from my PSU to power the watercooling unit. Well it turns out that it didn't have enough juice to actually run it all the way, but it fooled me by powering on the lights, etc. I had my finger on top of the waterblock when this happened, and the cpu eventually got hot enough to where I couldn't touch it, at which point I cut the power.

I quickly figured out what was wrong, corrected it and turned the power back on and viola! My new idle load temps are 37/41C. Great! System boots fine, everything seems to be working, no leaks. I decide to fire up prime95 to try and figure out how hot the cpu will get now under load. I decide to use the second test down, the one it says generates maximum heat. I start it running, throw on a dvd and I am about to open a beer when.... NO! Prime95 errors out, saying hardware failure. I check my CPU temps, 40C! I try running the other test, the memory one, that runs fine (ran it for a few hours straight). I then try the FPU test (first one) and that errors out.

So I tried upping the voltage, and it has remained unstable. I currently have worked up to 1.75 (from 1.65). Everything else seems stable, but two out of the 3 prime95 tests say I have a problem.

My questions are as follows:
1) Do you guys think I may have damaged my proc during my watercooling install (when it got to hot to touch). If so, will more voltage help get back my stability?
2) Since I only did one test on the memory when I declared the system stable (memory intensive torture test), is it possible that these FPU tests on the processor are failing because my memory still isn't performing where it should be?
3) Is there any way I can better (or independently of prime95) isolate which piece of hardware is failing?

Sorry if this is in the wrong forum... I am currently have a 50/50 split over which part I think is causing the issue.

Thanks,
-zbose

 

[c627627]

Desktop Bartons may need 1.85 or more Vcore for major overclocks.

Edit, wait, are you talking about running this at stock speeds?


Try relaxing RAM timings.

Post power supply brand, FSB speed (default 200?), MHz speed, case fans, case temperature?

 

[zbose]

Desktop Bartons may need 1.85 or more Vcore for major overclocks.

Edit, wait, are you talking about running this at stock speeds?


Try relaxing RAM timings.

Post power supply brand, FSB speed (default 200?), MHz speed, case fans, case temperature?

Guess I should have been more specific.

Stock speed, 200 fsb (2.1 Ghz).

Ram is supposed to run at 2-2-2-6, I guess I can try relaxing to eliminate it as a problem (or isolate it as the problem).

More hardware details:
Lian-LI PC-61 case
2 intake fans by harddrive bay,
1 outtake fan, 1 powersupply fan
average case temp is 25C
average socket temp 37C
average diode temp 37C
Antec Truepower 430 PSU
200 FSB (10.5 mult)
Barton 3000+ @ 1.75v right now
Mushkin 222 special (2x512) in dual channel, 2-2-2-11 @ 2.8v

So no overclocking yet (don't have a stable system!) but I am running agressive timings... of course thats why I bought the 222 mushkin stuff.

Does that help?

 

[zbose]

Update: prime95 has been running the first torture test (heavy FPU) for nearly 1 hour now with no error... Grr. This behavior is so incosistent.

 

[c627627]

You should be stable with your memory timings set to Automatic in BIOS.
...and sorry for misunderstanding you, people usually post overclocking problems in this section so I assumed your q was like that too.

 

[zbose]

So you are saying that if I relax my timings and that makes my system stable, then my ram is obviously the culprit?

 

[c627627]

Your RAM should run at default speeds and default timings stable providing you installed your setup (especially cooling) correctly.


Your mobo will run default timings if you set it to Auto in BIOS, don't set it manually at first if you want to test it.

 

[zbose]

Your RAM should run at default speeds and default timings stable providing you installed your setup (especially cooling) correctly.


Your mobo will run default timings if you set it to Auto in BIOS, don't set it manually at first if you want to test it.

So here is a new development. I was able to run the first torture test for about 2 hours, then I decided to cut it short and run memtest 86, this is at 2-2-2-11. It ran through one pass just fine. Then I set the bios to auto for ram timings and it chose 2-2-2-6. I just booted to windows and now I am trying that. But so far, it appears the ram is rock solid, as long as I have it 2.8v.

So could it be the CPU? Could it be something else?

 

[zbose]

Well, as per your suggestions, setting the timings back to its default (auto) seems to have made it stable. about 1.5 hours now with no error *fingers crossed*

-zb

 

[zbose]

Argh! After running Prime95 w/ in-place large ffts all evening while I was playing video games, watching tv (on the computer), recording tv, browsing, listening to music, etc it never once hiccuped. I then start a fresh iteration and go to bed.

I woke up this morning to find that it errored again after 1 hour 17 minutes!

I just yanked out one of my sticks of mushkin. I will see if it runs stable with just the one.

Im starting to get frustrated here. Any ideas?

PS- When I pulled the one stick of ram out, its heatsink was warm to the touch, not too hot to hold, but warm. How likely/possible is it that my ram is overhearting after an hour + of prime95? I am running it at 2.8v....

 

[Milspec]

Argh! After running Prime95 w/ in-place large ffts all evening while I was playing video games, watching tv (on the computer), recording tv, browsing, listening to music, etc it never once hiccuped. I then start a fresh iteration and go to bed.

I woke up this morning to find that it errored again after 1 hour 17 minutes!

I just yanked out one of my sticks of mushkin. I will see if it runs stable with just the one.

Im starting to get frustrated here. Any ideas?

PS- When I pulled the one stick of ram out, its heatsink was warm to the touch, not too hot to hold, but warm. How likely/possible is it that my ram is overhearting after an hour + of prime95? I am running it at 2.8v....

Can ram "overheat"? I didn't think that ram heatsinks were actually for the dissipation of heat... i thought they were just for spiffiness factor...

 

[zbose]

Nope, they are there to help dissapate heat! People will tell you to make sure you have good cooling when you overvolt your ram.

 

[c627627]

The system componenets are supposed to run stable only at default, automatic voltages and speeds. Everything else is overclocking and not guaranteed.

You need to have installed your system correctly too.

If you have doubts and have access to another system, simply put your RAM into another system and test it there, (rememeber: at default automatic settings and speeds, not overvolted and not with manually set timings.)

 

[zbose]

The system componenets are supposed to run stable only at default, automatic voltages and speeds. Everything else is overclocking and not guaranteed.

You need to have installed your system correctly too.

If you have doubts and have access to another system, simply put your RAM into another system and test it there, (rememeber: at default automatic settings and speeds, not overvolted and not with manually set timings.)

The problem is that the ram does NOT run stable at the automatically set timings (it chooses 2-2-2-6). It is mushkin RAM, the 222 special. It is supposed to be tested and hand picked to run stable at those timing and comes from the factory like that. I would have to manually set the timings slower if I wanted to run 3-3-3-8 or something.

At 2.8v the ram passes memtest86 and any other test I have tried. Also, I have been running one stick for close to 3 hours now doing in-place large FFTs. I would replace it with something else to test, but unfortunately, I only have one machine.

I found this thread: http://www.mersenneforum.org/showthread.php?t=1588
and I am now running the newest version (2.3.8). That thread indicates that for 1024 MB setups, there may be prime95 memory allocation problems. When I get home tonight, I will throw the second stick back in and re-try it, as it looks to be nice and stable with 512.

 

[c627627]

Try this, set your system to all default automatic voltages and speeds.

With only one RAM stick, boot from a memtest86 bootable floppy:

http://www.memtest86.com/

Memtest86 cannot be executed under Windows.
It needs to be loaded from a floppy disk.

Double click on install.bat to make a bootable Memtest86 floppy disk.

Note that after the boot floppy has been created you will not be able to read the floppy from Windows. This is normal.


Then if you pass, put the other RAM sticks in separately to see if just one of them is bad or what....

 

[zbose]

Try this, set your system to all default automatic voltages and speeds.

With only one RAM stick, boot from a memtest86 bootable floppy:

then I decided to cut it short and run memtest 86, this is at 2-2-2-11. It ran through one pass just fine.

Then if you pass, put the other RAM sticks in separately to see if just one of them is bad or what....

So it runs memtest86 with both sticks in dual channel. Unless I need to loop it for longer or whatever to bring out the errors, memtest86 is not complaining.

Also, prime95 torture test (inplace ffts) has been going on for about 4 hours now with out complaint (with one stick of ram). I think that it may be the issue mentioned in my previous post about prime95 having a bug for >512 memory setups.

 

[Sniper.nkc]

Memtest86 is for testing your ram. If you can achieve the rated speed without errors, then your ram is good. If you get errors, then your ram is bad. You need to solve this and not include this as a CPU issue.

If Memtest86 pass. Then go ahead and set-up last know good setting. If this setting produces an error, then you have a problem. I noticed your CPU was hot at one point. the heat may have weaken/damaged your processor. This may be your culprit, but first you must rule out all possibilities. You may have to start at the beginning to know where your OC is peaked at. Good luck.

 

[zbose]

Memtest86 is for testing your ram. If you can acheive the rated speed without errors, then your ram is good. If you get errors, then your ram is bad. You need to solve this and not include this as a CPU issue.

If Memtest86 pass. Then go ahead and set-up last know good setting. If this setting produces an error, then you have a problem. I noticed your CPU was hot at one point. the heat may have weaken/damaged your processor. This may be your culprit, but first you must rule out all possibilities. You may have to start at the beginning to know where your OC is peaked at. Good luck.

That is what I was wondering. Did I damage my CPU?

Memtest86 DOES pass at the same settings prime95 fails at. So i have basically ruled out that my memory is bad. Which leaves me with my CPU possibly being bad...

 

[Sniper.nkc]

Don't give up yet. Many things could have happened that you may not remembered when upgrading your rig. Frustration is your worst enemy, so take your time and double check everything. Check your temps. If you have exhausted all possibilities, then go ahead and built your rig back to original configuration with air-cooling and good known settings and verify. If you still get an error, then your CPU is probably your problem since you did expeirenced an significant event (hot processor).

You may want to get a new processor to test that theory, and if the new processor did not fix your problem, then you can return it and get your money back too. No lost there!

 

[zbose]

Thanks for the support! I know this is all just part of the game.

I am happy to report though that Prime95 (2.38) has now been stable for 6+ hours... Things are looking up! I can't bring myself to believe that the heat I during my screwup would be enough to hurt the processor. If it had been coffee, I would have been able to drink it! Just too hot to touch.